Shinya Kurosaki

Analysis of Blend Irregularity in Yarns Using Image Processing Part III: Evaluation of Blend Irregularity by Line Sense and Its Application to Actual Blended Yarns

In Parts I and II of this series, we discussed the precision and efficacy of an image processing system for evaluating blend irregularity in blended yarns using three kinds of model yams, and reported on the application of that system to actual blended yams. That system involved converting original images into trivalued images (black fibers, white fibers, and gray background), which enabled quantitative evaluation of the blend irregularity on the surface of a short length of yarn (about 4 mm). However, that system required threshold values for converting original images into trivalued images, and also needed a considerable amount of processing time and experience. This study deals with a new system to evaluate blend irregularities by measuring line sense (a series of intensity values) around the center line along a yarn axis without converting the original image and by analyzing the distribution of the intensity values. This method saves processing time and provides a real-time evaluation of the blend state ( fiber arrangement). We have studied the feasibility of evaluating blend irregu larities by means of measuring line sense with this system using two kinds of model yams, where the numbers of component fibers, blend ratios, and blend states are known. We also discuss application of the system to actual blended yams.

Analysis of Blend Irregularity in Yarns Using Image Processing: Part I: Fundamental Investigation of Model Yarns

In this research, a computer image processing system is used to evaluate the blend state (fiber arrangement) or uneven distribution in blended yarns. Part I presents an investigation of the precision and effectiveness of the system using three kinds of model yarns, whose numbers of component filaments, blend ratios, and blend states have been predetermined. The method of image input, various image processing techniques, and a statistical analysis are examined based on the model yarns. The results show clearly that with this system, the diameter and area of one fiber can be measured, the blend ratio of black and white fibers in a short sample length can be accurately determined, and the size of a cluster can be quantitatively evaluated.

Analysis of Blend Irregularity in Yarns Using Image Processing: Part II: Applying the System to Actual Blended Yarns

Image processing has been applied to actual blended yarns made from wool/wool and wool/mohair, and the blend state (fiber arrangement) of the yarns analyzed. Yarn about 4 mm long (310 pixels) was processed in each image. The projected area of white and black fibers on the yarn surfaces, the blend ratios, and the sizes of clusters were determined. Further, when several scores of yarn images laid in a row were put into the system, blend irregularity curves could be obtained. From these curves, variance-length curves of blend irregularity were calculated, and a total variance, average values, etc., were determined. The analysis showed that the wool/wool yarn was well blended compared with the wool/mohair, because the former was smaller in blend ratio variance than the latter, was also smaller in cluster size and its variance, and further was smaller in the total variance of the variance-length curve for blend irregularity. With this system, such differences in the blend state of an actual blended yarn, which can hardly be distinguished with the naked eye, can be estimated.